The collected works of tillin9 - Page 1

PM sent.

EDIT: While I PMed yesturday, the message seems stuck in my Outbox.

Anyway... I'm in Middletown (~30 min away I believe) and can drive over this weekend to pick it up. However, I wanted to check if it was an Octane or Octane2, and what condition the skins were in beforehand.

Thanks in advance.

If the skins aren't too badly damaged I can always do my restoration trick. I care more about the internals as I might want to upgrade the processor/ RAM, so that's good to hear. Video being IMAPCT is perfect, as Linux can't do VPro. I do sort of want everything, with the V6 being the lowest on priorities.

I happen to have a Fibre issue at work that this guy could really help on debugging. I have plenty of cables cables and GBICs or miniGBICs (if the card needs them), so that's not an issue. The model of the fiber card, if its in a PCI shoehorn, or if it needs an SGI DB9 to optical (MIA) would be useful.

Assuming I end up with it, I won't need the Fibre array after about a week's worth of testing (mainly to debug the card so I know it isn''t a problem on my end when I connect to the University computational SAN), so I'd be willing to pass it on to someone who does. I may need to hold onto the card depending on model and other issues. If its a X9-FC-2POPT-B I'd be really happy.

Still interested. As I said above, if I do get it, I'll gladly pass on the array once I'm done with it (a week tops). Shipping to Texas will probably be expensive, though.

PM sent.

I2C is the protocol used by sensors to let the BIOS know things like temps, fan speeds, and supply voltages.

The

vs.

looks like the issue to me. You seem to have a problem with your PSU adaptor (i.e. not all rails are getting power).

If you're sure the wiring is okay, you may have damaged your voltage monitoring IC.

Finally, since the video card does have an I2C connection to monitor fans and voltage on most newer cards. It may be on the Fuel this system is integrated to the central I2C controller and not on a secondary one on the XIO bus, so the lack of a video card means an I2C error. Or the PROM is smart enough to sense the secondary chip. You may be able to disbale this somehow.

I must admit I first learned how to program in BASIC. When I was little the first thing I wrote (well, that I remember) was a game in text mode graphics on an 8088 where you played the Dinosaucers figthing whoever they did in the cartoon (it was a long, long time ago). While I think its best to learn C before diving into other languages like assembly, C++, PHP, etc. if someone is too young to learn C then teaching BASIC might be a good idea. You get a lot of the math and simple programing concepts without all the complexities of pointers, scope, datatypes, etc. that are critical to being able to use C for anything more than hello world.

Plus BASIC->PBASIC (Parallax Basic Stamp language) is a great way to learn about electronics and microcontrollers without having to know assembly if you ever decide to go that route. That's not to say one shouldn't learn assembly, my knowledge of assembly has really helped my C (and other language) coding so I think that its a vital part of any programer's study which is sadly in contradiction to the Java only curriculum of most CS courses (at least in the US).

If you have an old PC, the BASIC interpreters IBM made were quite nice (and the manuals very well done). Windows 3.1 era machines came with Microsoft's QBasic (its on Windows 95 and 98 CDs even though its not installed by default). I'm fairly sure that DOS-box could run all this in emulation on a UNIX or XP machine if you don't have the hardware.

Anybody try a newer ATIon the O2 with X.org > 7.1? I got my legacy IO API working, and X.org sees the card, but still pukes. I haven't had time to play around with newer X.org builds, but have been constantly told they're re-arranging their PCI subsystem.

If anybody has some low-level graphics or even x86 and MIPS assembly (I'm seriously lacking on the x86 side) programming experience, I'd love for them to take a look at the Int10 + ATI + Legacy IO path, its very possible that the Int10 routine needs some MIPS/ O2 specific code. Int10 converts x86 video BIOS to platform assembly to POST the card which is a non-trivial task. However, I ported the underlying kernel API that lets X.org see it as a video card (find registers, pull the BIOS rom, etc.). See: viewtopic.php?f=8&t=13680&start=60

Note, with a 600 Mhz processor compiled with the proper RM7K optimizations, the O2 makes a half-decent desktop even without 2D acceleration. With it, It would be on par with a high end PIII, maybe better with ICE patches and X and KDE compiled with optimizations.

Also, from the Linux-MIPS mailing list, somebody just got a patch to get intelligent PCI extenders working on the O2. I.e. you plug one of those into the single PCI slot and it provides the necessary bridges to get 3 PCI slots. A little hacking with one of those, a video card, and a small USB PCI could give you all the IO you'd need.

zahal wrote: BTW, What's the R12ka optimized port status?

As far as I know Linux will run, but crash after 5min-hour because of the R10K cache issues. Octanes have hardware to prevent corruption and BSD uses a whole other set of DMA routines which is really what prevents Linux from being stable. NetBSD and OpenBSD should have R10K and R12K O2 ports, but the work done for those OSes don't directly apply to Linux. The best info. on the subject is here: http://www.total-knowledge.com/progs/mi ... ssue.shtml The same site has some patches, but work is slow...

As far as Linux/ Irix on a single drive, that's a walk in the park. You just need to use arcload (or the more complex arcboot) and supply the right partion numbers and kernel names. Tell it to boot one by default and if you want another OS, interrupt the boot sequence and type boot Linux or boot Irix (or whatever you name them). I do this all the time to control what kernel I load and whether to load my Debian or Gentoo partitions. I actually haven't tried Linux/ Irix dual boot since I have an R12K O2 that runs Irix and I use my 600 Mhz one purely for Linux, however, I know its possible. You may need a larger hard drive (or to resize your volume header since its advantageous to put both kernels, arcload, and config files there) to do this. I'd highly suggest trying the Gentoo SGI Live CD first though.

If you want to try with arcload, I can help. Arcboot is a whole other issue. Gentoo uses this instead (by default), and it doesn't seem to properly overwrite the OSLoadPartition PROM setting so you either need to change PROM variables before booting (annoying) or do some major trickery to get Irix to load. See this (rather confusing) thread: http://forums.gentoo.org/viewtopic-t-54 ... +boot.html

As far as power, I'm okay since my 600 Mhz processor draws < 3 Watts (it probably could run without a heat sink). Trying to add a modern video card on an R12K probably isn't the best idea, but for heat issues rather than power. Open the case of your running R12K and feel the giant heatsink after 30 minutes, rather warm is an understatement.

The fastest PCI ATI card you can get (at this time) is an X1650 and still only draws around 30 Watts full load, which is the max a standards compliant PCI slot can provide. I know the BFG PCI says 250 Watt, but I think this is more of a cya number since most power supplies inflate their wattage. Also, the PCI board doesn't have a power connector so it must be under 30 watt, like the X1650. As far as the numbers on the box go, I've run P4s off high quality 150 Watt power supplies when the box said 300 minimum (and drew a ton of flak for that) but using a Kill-A-Watt measured the real peak load of that system to be closer to 135 Watts. Since an O2 can also accept a secondary video card and video capture board, I'm fairly sure power supply isn't an issue unless you really overdue it. Your PCI extender may need to tap into the power supply directly if you plan on using multiple bus-powered USB devices, but I'm sure its doable, even on an R12K.

Finally. you won't think this from the monster heatsink and how hot it gets, but the R12K @ 400 Mhz is only a 20 Watt processor (according to wikipedia). However, I'm fairly sure about this number since Hennessy and Patterson mention the R14K @ 600 is 17 Watt and that number is also in the Wikipedia chart. That means 150 watts for all the other components in an O2.

Anyway... sorry for being long.

Seems silly, but why not put the table inside the <form> tag? You can then either have 1 submit button or 1 per line and then do server-side parsing to let the user know more data is needed (or not, thinking like an online order form). The only thing is I'm assuming you can use HTML 4? If you just didn't think of this sorry, but having to use forms only inside of tables seems like a Netscape 3.0 support decision or something to me.

Using pure HTML 4.0 and CGI I'm 99% sure its not possible. I mean if you have multiple forms per page, and submit one, does the server get any info on the other form's attributes? No, its like they're not even there. This is an HTML issue, not a server-side issue. If you can go XML 2.0, XForms lets you do data binding, so two Xforms can share the same attributes with some work since the document has a singular namespace.

If you have to stay HTML, a hack would be to shadow every form's input elements in every other form and use Javascript to toggle all the hidden elements when one real element is toggled. You may be able to accomplish the same via CSS 2.0+ since CSS lets you toggle names elements properties - this is how you do CSS only rollovers and menus, essentially you add text to a hidden element that is sent, then parse it on the server.

Honestly, though, I think it would an order of magnitude easier to just put them all in one form even if it means modifying the legacy code. I mean how hard it is to add a server side variable to not echo the <form></form> tags per line (I'm assuming your code generates each HTML form and checkbox via a function in a template.) and stick one around the whole nested table structure. Maybe a few lines to change the naming scheme, and you're done as each submit now submits the whole form.

Wow, that cryptomemory is some neat stuff. My first instinct was that you could get a 2K SEEPROM and fake it, but I don't think thats possible since the chip doesn't use I2C protocol. A small PIC probably could work as an easier to implement stand-in.

Assuming SGI didn't just use it as PROM and uses the encrypted ATR protocol, you'd need to at least need to log the datastream sent to the module and do some cracking to figure out what the right bits are to send. The encryption seems to be 64-bit, which could be rather painful (DES is only 56) depending on the implementation.

Probably best to call the company and ask for relicensing, if possible.

Honestly, I think its cool, but kinda insane. I briefly thought about doing something similar as I have a stack of 2Gbit SFPs leftover from work and a bunch of PCIX fiber cards which are supported under IRIX/Linux/BSD. The main thing which stopped me is that electricity here is very expensive and a full-fledged computer (as a gateway/ firewall/ web server), with a fibre switch and wireless access point consumes a whole lot more energy than my WRT54GS. The heat/ noise factored in there a bit as well.

I know your network has had issues, but trying to "standardized" on fibre is a bit of an oxymoron. Fibre is a very proprietary media and without long distances, it offers little benefit to gigabit often at much higher cost.

I know it lacks the cool factor, but if I were you I'd sell a good portion of that fibre and purchase a good 16-24 port 10/100/1000 fanless gigabit switch with a 1000ft spool of CAT6, a crimper, and bag of RJ45 connectors and maybe some 100Mbit network cards. Ideally you can use expansion cards to make your 10Base2 machines into 10Base-T ones but media convererts do exist and are cheaper than ones to fibre.

Have any LC-LC? I'll trade a bunch of roughly 1000ft CAT6 lengths for a few LC-LCs. We're actually in the process (we'll its somewhat constant since we routinely add new machines) of adding new cables for the research clusters. University IT gave us all the leftover cable they were going to throw out a few months ago and we have more than we'll ever need. I'm sure I could get more very easily as they are horribly wasteful. I can't find any links for anyplace that sells bulk cable like that, but it comes in on trucks via 5ft high plywood spools.

Anyway... sorry to be a downer.

Fibre is cool and can be fun to play with. I liked getting to use 2Gbit links between our main cluster switches at work. However, while all was fun with the same model Linksys switches, trying to get them to interface with the University's Cisco gear was absolute hell. I'm not saying get rid of all the fibre, a single high speed link between two machines like your main workstation and fireserver is fine. Just realize its not something really possible to make a full home network out of without a lot of pain and money.

Also, the sole benefit of using fibre at work besides uping our headroom (we haven't yet maxed out gigabit) was to decrease latency, which it did for longer hops on the network by roughly 60%.

As far as the Mac Intel vs. PPC issue, the only time I've ever found the G5s to be faster is when the software uses Altvec optimizations and hasn't been re-written for SSE. This was a big issue with Final Cut before my old boss got her new software, that the new Intel Macs took 2 minutes longer to compress things than the old G5s. With proper software support, most encoding is now sped up by > 2x.

The dual G5s pump out so much heat, especially when used heavily for video work, I'm not surprised by the "tempermental" quality of the quads.

Finally, a 30" display isn't for everyone. I tried the 3007wfp and couldn't use it, text was too small if I sat at a comfortable distance and I had to move around too much if I got close enough to see. I found the 24" a much better fit, plus it was about $1600 cheaper since I would of had to repace my video card also as my aging ATI can't do dual link DVI. You can get a 24" E248WFP off of Dell for around $410 but its crap. My boss has one and though he loves it, I think the image quality is exceedingly poor. Basically the image head on looks like my 2407wfp if I try viewing at 170 degrees, a little fuzzy, poor gamut, poor contrast, poor brightness. I got my 2407wfp for $540 and think the extra money was worth it. You might want to try for a 2405wfp since its basically the same panel (just doesn't support 1080p via component) and you can get it cheaper.

I've managed to upgrade my Octane2 and while Linux runs like a dream on it, sadly I think I need to re-install and/ or upgrade IRIX to deal with the newer processor. While I know I could try to do it via network, I think a CD drive would make this task easier and help a lot with other software installation and just day-to-day use. A couple questions:

1) My question is what type should I get? I assume any 68 pin and 50 pin w/ adaptor will probably work but always helps to double check.

2) Did SGI ever make any themselves? I won't mind a matching or branded enclosure.

3) If not, anyone know a model that's reasonably quiet? I used one on someone else's Octane which made almost as much noise as the Octane.

Sorry, I guess I didn't quite phrase this quite right. I'm more interested in the enclosure. I have a spare Plextor 12/10/32 known to work with SGIs, and if all else fails, I do have a spare O2 CD-ROM I could use.

I'm surprised I never saw a picture of the SGI branded one. Anybody who has one mind snapping one?

Neko, thanks for the pic. Looks a little more plain that I expected. I happen to like my Zip650 on top of my Indigo (although connected to my PC) since its a matching purple, and My Jaz next to my Octane as they're both green (I'm probably the only one with all the insides of an Octane2 in a set of green skins).

Naw, I still have an old dirty 6.5.11 I didn't install. I have 6.5.22 overlays, but never got around to using them on the Octane.

Can I do an install over network from an O2's CD? That would work.

Care to share how it was configured? Or is this just a normal network install?

Besides iSCSI, I can't think of a good way to share the device.

Um... Why not just put a copper GBIC into a regular Fibre hub? Something like this: http://cgi.ebay.com/New-Finisar-1000BAS ... dZViewItem

Quad GigE PCIX/PCIe cards do exist (and actually depending upon driver make bonding easier too), making it only need 6 expansion slots. Some of the motherboards with those slots also have built-in U320, I'm thinking of the Tyan Thunder S2676 (3 PCIX, one PCIe, 2 PCI which is the heart of one of my servers at work) which takes it down to 5, so that means 2 quad GigE (say PCIX), FDDI, Thin Net, and Token Ring PCI. Doable, but not easy or cheap. Of course one could also imagine the server being some form of Origin or more realistically not having 3 bonded GigE lanes to your fibre channel or 3 bonded 100 Mbit to your router.

Seriously, why the Linux hate? Yes, I'll third that the Octane's hardware works much better under IRIX and that any of the really unique stuff like video capture and accessories probably don't. However, the original question was has anyone gotten Linux to work on their Octane. The answer is yes, I have. Kumba (the Gentoo-MIPS lead) and Frapazoid are two other forum members who I know have as well.

The Octane isn't quite as crippled under Linux as some posters have suggested. Impact based cards have some degree of X acceleration. The VPro cards don't, however. The audio works, onboard SCSI works, the Octane tape drive works, and PCI shoehorns and card cadges work (and support a large amount of hardware IRIX doesn't like USB cards). I'm not 100% sure on the status of the XIO peripherals, some cards work (anything that internally uses a PCI bridge) including SCSI, Gigabit, and Fibre Channel. The only other thing I remember testing was the SCSI floppy, which didn't work. (Actually caused constant printks so I had to pull it).

Somewhat unfortunately, all of the support is currently only in Gentoo. I highly suggest the Gentoo SGI Live CD. If you have a SCSI CD-ROM, just burn it and boot. If you like what you see, you can go from there. http://dev.gentoo.org/~kumba/mips/livecd/x-rc6/ Actually read the readme since its not a normal ISO image.

Getting the kernel patches into Linux-MIPS and mainline is a goal of one of the developers on the mailing list who recently added a lot of SGI support (audio drivers for O2, Indy, and Indigo 2, among other things). Also, I'd like to point out that 2.6.27/28 will likely make the VPro issue moot as Linux will get kernel mode setting. That means Radeon X1550s (or whatever's the best PC PCI video card with Linux drivers around) for our SGIs. I'm fairly sure an X1550 is better than a V12 for the vast majority of video tasks.

Anyway... the point is that Linux support for SGIs is being worked on and will get better as time goes on.

Sorry, I think I overdid it.

Yea, even though I have a V12, I was looking around for an MXE for awhile for the sole fact that its the fastest card with Linux acceleration.

As far as the kernel mode setting, I haven't gotten that working on an SGI. Actually, the only really working support (as in testable with a released distro) right now is Intel's integrated graphics. If Intel actually releases Larabee, that should work too. Open source radeon support is going to be included in kernel 2.6.27 or .28 (I know the devs. that are working on the ATI side - and have told them about my plans to put a card in an O2) but the current stable linux release is 2.6.25 so we're a few months off. When I get it working, I'll let everyone here know.

@Chatuser - Yea, Debian's kernel is based on either Linux-MIPS or the mainline tree and hence won't work on an Octane. As I said, the necessary patches are in Gentoo's repositories. You can use a Gentoo kernel with Debian, but actually getting it on the Octane would be a little tricky. I'd recommend doing something like installing on an O2 or Indy (which Debian supports very well) then compiling a Gentoo kernel for the Debian disk and transplanting it into your Octane.

@pip - Seems like an arcload issue. I also couldn't get arcload (Gentoo's choice of bootloader) working, so I used arcboot (Debian's choice of bootloader) instead. Arcload is technically better, but more difficult to configure. The config file needs lots of nested { and } at the right places. At the time the nice wiki and help entries didn't exist, so you might have more luck now using the live cd and fiddling with the config.

You're not the first person to doubt my statements, but a big thanks for bugging the devs yourself and making a nice article for me to link to in the future. After your first responses I was going to go dig up all the old irc / forum posts where I hashed it out the first time, but now I don't have to.

What really surprises me is that more alt-arch people don't know / care about kernel modesetting. I'd think there'd be tons of embedded MIPS / ARM people who program devices with PCI buses (you'd be surprised that this is the majority, and most of the rest have PCIe 1x) who would love this. Just stick in a PC video card and have a low-end desktop environment for debugging / development instead of messing around with serial ports and JTAG.

While AtomBIOS and AMD/ATI's new open source policy makes this a lot easier for radeons, there is no reason why kernel modesetting won't work for all cards. Intel does it themselves, and I'd actually think older cards would have a lot simpler sequence than even an AtomBIOS parsing radeon routine. That leaves Nvidia to be reverse-engineered, and the Nouveau devs seem to think they can do it. http://www.phoronix.com/scan.php?page=n ... &px=NjU0Mg

Also note that the radeonhd camp doesn't like AtomBIOS and bit bangs all the registers manually (although the higher ups got a little annoyed since it took much longer to support cards than the radeon folks which used AtomBIOS). AtomBIOS is just a nice interface for register table lookup allowing the same initialization code to reused. It doesn't actually initialize the card itself.

As far as to why, I got into Linux on SGIs since I needed to teach and intro assembly programming course. Most courses either use embedded MIPS devices which make learning very hard because they don't have enough power / memory to compile things themselves or try teaching x86 and confuse the hell out of everyone. In the first case, 9/10 times the problems aren't with understanding the assembly but with getting the program on the embedded device (cross-compiling, bricking the device while flashing) or being able to debug it (JTAG) when something goes wrong. I could just take the proper time to teach everyone how to do all that, but the department and university really look down on skills based classes. One intro to UNIX class was canceled for and I quote "Being too practical." Thus I got a couple of O2s, installed gcc, and was able to stick to Patterson and Hennessy.